On Thu, Dec 16, 2004 at 08:11:44AM +0000, Jeff Zucker wrote:
> Tim Bunce wrote:
>
> >After the "dbi:AnyData:"; everything should be in "name=value;" format.
> >So "dbi:AnyData:format=XML;table=Test;file=test.xml";
>
> Ok, I thought one additional param was allowed, as long as I parse it
> out, but I'll change it. Some other DBDs do this with the database name,
> (e.g. ODBC with a named ODBC DSN) so should they also change this behaviour?
In the sense that dbi:ODBC:db=foo and dbi:ODBC:database=foo should
be supported in addition to dbi:ODBC:foo, yes, that would be good.
> >>b. my $flags = {format=>'XML',table=>'Test',file=>'test.xml'};
> >> $dbh=DBI->connect('dbi:AnyData:','','',$flags);
> >
> >You've called it $flags but it's really \%attr
>
> Yes, I should remember to call it that in the docs.
>
> >and the rules for
> >attr are that driver private attributes should have a driver-specific
> >prefix.
>
> Yes, I know, and none get stored without prefixes. Since these are
> table-specific attributes, they are not stored directly in the database
> handle, rather under the $dbh->{ad_tables}->{$tablename} hash.
But in this specific example they are defaults stored at the $dbh level.
> I delete
> all of the lower-case attributes out of the \%attr in connect() and then
> feed them to ad_catalog() or ad_import() as appropriate (which put them
> in the ad_tables hash after munging). The few attributes that are not
> table-specific, I add the prefix before storing them back in $dbh. So
> all driver-private database handle attributes *do* have the prefix as
> soon as connect is done.
Fair enough.
> Should I enforce the prefix in \%attr even
> though I just strip it out and do something else with the attributes?
Yes, at least in theory. Remember that \%attr is meant to be able
to contain attributes for many different drivers without any of
them clashing. That's what the prefix is for.
> On another topic: I am about to release a rather major interface change
> to Text::CSV_XS. Obviously it's not directly DBI but it does impact
> those who use Text::CSV_XS with other DBDs, should I post an RFC here,
> or in dbi-users, or not on the DBI lists at all?
A note on dbi-users seems reasonable to me.
Tim.
